Next steps for Crypt-SSLeay
My primary goal for the 0.58 release was to understand and fix as much as I could without breaking anything. Occasionally, that meant leaving certain blocks of code in place in
Makefile.PL even when I was not too sure it was necessary to do so. I did try to improve version detection, fix a couple of things and factor out code to subroutines, but I left much intact.
Two new bug reports, RT #61249: Over-rigid library use on Win32 prevents dynamic libraries and RT #61324: Broken for *static* (also) openssl builds on Win32 highlight the need to rethink the organization of
Makefile.PL for this module.
This is a relatively straightforward task, but seems to require some upfront thinking to accommodate various permutations such as independently installed mingw versus mingw bundled with Strawberry Perl versus mingw bundled installed with ActiveState's
ppm versus MS VC along with ActiveState Perl etc as well as dynamic versus static linking of OpenSSL.
Matters are made slightly more complicated when I consider threading which is the topic of RT #41007: [PATCH} ithreads support. As noted in OpenSSL threads documentation,
crypto/threads/mttest.c includes sample code I can incorporate into
SSLeay.xs. I would like to include threading support for both Linux and Windows platforms in the next release and have a
Makefile.PL that handles all permutations correctly.
That should happen in a couple of weeks. If there are no problems with 0.58_02, I will release 0.59 soon thereafter. I will add threading support in subsequent development releases.
In the mean time, I would appreciate input on how to write tests for threading support. In particular, short scripts that reliably fail in the absence of threading support would be of great value ;-)